Storage: Keeping data in the frame

Whether or not it can be attributed to the explosion in demand for data warehousing is a moot point. But either way, mainframe storage ? once a topic guaranteed to make the eyes of any IT manager glaze over ? is sexy again.

?Mainframe storage has evolved from a Cinderella technology into one of the key issues facing IT directors today,? says Sharon Forder, Prism Solutions? international marketing manager.

In a world where, increasingly, only the fittest survive, organisations are recognising that knowledge is power. This has had a far-reaching impact on the storage market as organisations today acquire an appetite for accessing the data they have available, not merely as disparate sound bites, but as complete enterprise-wide information resources.

Mainframe storage has become a strategic asset as corporates wake up to the fact that their data is their company. They are evolving storage strategies to reflect this. Storage developer CLARiiON?s European director Mike Brown says: ?With the evolution of thin clients on the desktop, centralised management of data across multiple computing platforms is becoming critical.?

Brian Hayes, UK country manager for Legato Systems, adds: ?All offices continually create and receive data, which needs to be filed and stored for a specific period for reference and legal requirements. The mainframe is vital as a storage medium, though it is often only at moments of disaster that the mainframe?s vital role is revealed and the true strength of the security of the system is demonstrated.?

It?s generally agreed then that the tide has turned in favour of the mainframe, which has been enjoying a resurgence in its popularity for the last three years.

John Campbell, principal at consultancy firm CITL, says: ?Before that, everyone said the mainframe was dead and we should jump on the Unix bandwagon. Now more and more processing, whether it?s NT, Unix or the mainframe, is being consolidated ? and so is storage.?

Storage vendors are responding to the trend with solutions that share data between mainframe and Unix systems. Campbell explains: ?You can carve up a single pool of storage between the various servers. You can attach that storage to both mainframe and non-mainframe systems, and use the features of a mainframe as a way to back up and recover that data.?

For example, EMC is starting to blur the distinctions between storage devices for mainframe, AS/400, Unix, NT and other open systems as its Symmetrix system consolidates information from each platform.

EMC?s UK and Ireland marketing director, Mike Maunder, says: ?As we reach the end of the decade, processor distinctions will be replaced by one term ? enterprise storage. One storage system will be able to provide a complete view of all of an organisation?s information, whatever the host processor.?

IT managers have always been able to mix and match mainframe storage systems, but the technology is changing. It is moving on from tape to RAID (redundant arrays of inexpensive disks) technology.

Tape is still complementary to RAID as a way of taking offline backups, but companies with heavy mainframe investment sites have used RAID technology primarily because of its high availability and performance for mission-critical applications.

Campell says: ?RAID is more reliable than tape and the older single large electronic disk (SLED) devices [which the technology is replacing]. Those with RAID usually have to think hard to remember when they last had a media failure.?

As a way to ensure competitive pricing, most mainframe shops now adhere to a dual supplier policy. They buy their storage from a combination of DASD suppliers, IBM, HDS, or HDS and EMC, most of whom have RAID offerings.

In the past 12 months, DASD vendors have started to offer functionality which can support year 2000 testing or disaster recovery.

Take DASD mirroring. It is possible to take a snapshot of the production data and copy it on to another disk locally without closing down the production system as you take the copy.

There is much debate about the relative merits of SCSI and fibre channel technology supporting storage on the network. Roger Cox, director of array marketing for Adaptec?s mission-critical sub-systems division, says: ?Fibre channel has exciting potential, but there is still plenty of mileage in SCSI.?

The quality of service promised by fibre channel, particularly across long distances, does lend itself to multi-site mainframe storage applications. Theoretically, fibre channel can support over 100 devices with transfer rates of 100Mbits/s over a distance of 10km. But there are cost concerns. Cox adds: ?Fibre channel is a high-end solution, but the current excellent performance of SCSI must be considered.?

The real issues around mainframe storage relate to data access, management, protection and integration. Data must be located where those who need access can get it, storage costs need to managed and data integrity maintained, and protection policies applicable to the data stored implemented.

James Whitemore, Sun?s storage marketing and business development manager, says: ?To achieve this, IT managers have to look beyond the boundaries of traditional mainframe storage capabilities and use technologies such as data sharing, so that only one copy of any data file is held ? one version of the truth ? to which all users have access, regardless of platform.?

Manchester takes the broad view

MANCHESTER Computing (MC) offers computing and networking services for the main University of Manchester and its Institute of Science and Technology.

The computing facility?s policy is to have the right product for the right application, so it uses a broad range of equipment from PCs to supercomputers.

Most of the university?s academic computing is performed on distributed workstations or PCs. Around 8,000 desktops mostly run Windows 3.x or Windows 95 on the network. The PC infrastructure is based around 250 Novell NetWare file servers, 49 of which are supported by Legato Systems? NetWorker. The aim is to provide public clusters across the campus, located in areas such as in libraries, complemented by departmental clusters for each faculty.

Local high-performance computing facilities are provided by three IBM Power Series 6000/590 processors and a Cray EL98 Dual Processor.

Manchester Computing also provides a number of IT services to higher education institutes, including:

l A Vector Supercomputing service based on a Fujitsu VPX240/10 for research councils l A Datasets Service based on a Sun CS6400 Superserver on behalf of a higher education funding committee

l An Online Public Access Catalogue Service comprising records for some of the largest university research libraries in the UK and Ireland

l IT services for a research programme on electronic journals involving 20 publishers and 10 higher education institutes.

Other services managed by MC include a centralised University computing help desk system on a Sun Ultra 1, and a client/server combination of a Sun Sparc Server 20 and a Sparc Server 1000, which run a library management system for the John Rylands library.

Manchester Computing used to support six separate file store backup schemes as part of a centrally provided storage service for email file servers and centrally managed workstations. But in 1996, several issues surrounding the backup facilities started causing concern.

Dave Haworth, head of the computing unit?s distributed facilities operation, explains: ?As each functional area operated with different software to a variety of hardware platforms, we needed to ensure a large range of expertise could be provided. Our team was totally overloaded, but we needed to control IT support within the staffing levels available, which became increasingly difficult.?

Local backup used to be semi-manual. The media had to be periodically reloaded so backups could not be taken over weekends. This raised the spectre of how secure new data could be over these periods. The reliability of MC?s equipment was also a concern, so MC had to keep permanent spares for the tape equipment. Because storage requirements for the national based services on the Fujitsu and Sun CS6400 grew rapidly, MC?s Memorex automatic tape library had to run at full capacity.

?It would have been possible to upgrade the device, but the cost of the upgrade and the age of the equipment made this inappropriate,? says Haworth.

Demand for email also grew rapidly and started outstripping the capacity of MC?s DAT autoloader backup. Even though MC more than doubled the capacity of its email file server disk in early 1996, further storage equipment was necessary to support it.

?It became apparent that we needed to provide centrally managed, fully integrated data storage facilities using automated systems,? says Haworth.

?We also needed a system capable of substantial expansion to support the constant growth in our campus-wide backup capacity requirements.?

Enter NetWorker for Solaris, Legato?s autoloader and archive modules, and a StorageTek 9710 DLT based robotic library which initially provides 5Tb of data capacity.

Each of the four DLT4000 tape drives can store 20Gb of data (typically up to 40Gb compressed). The StorageTek 9710 is attached to an Axil 320 Sparc server running Solaris and connected to the LAN via an FDDI interface.

This solution rationalises administration effort, freeing up operations and system staff to develop new systems, rather than manually backing up data. Some users have also been empowered to manage their own backup, with user representatives instigating data recovery, giving further productivity gains for the operations team.